Accuracy of computer-calculated andmanual QRS duration assessments: Clinical implications to select candidates for cardiac resynchronization therapy

Background: QRS duration (QRSD) plays a key role in the field of cardiac resynchronization therapy (CRT). Computer-calculated QRSD assessments are widely used, however inter-manufacturer differences have not been investigated in CRT candidates. Methods: QRSD was assessed in 377 digitally stored ECGs: 139 narrow QRS, 140 LBBB and 98 ventricular paced ECGs. Manual QRSD was measured as global QRSD, using digital calipers, by two independent observers. Computer-calculated QRSD was assessed by Marquette 12SL (GE Healthcare, Waukesha, WI, USA) and SEMA3 (Schiller, Baar, Switzerland). Results: Inter-manufacturer differences of computer-calculated QRSD assessments vary among different QRS morphologies: narrow QRSD: 4 [2-9] ms (median [IQR]), p = 0.010; LBBB QRSD: 7 [2-10] ms, p = 0.003 and paced QRSD: 13 [6-18] ms, p = 0.007. Interobserver differences of manual QRSD assessments measured: narrow QRSD: 4 [2-6] ms, p = non-significant; LBBB QRSD: 6 [3-12] ms, p = 0.006; paced QRSD: 8 [4-18] ms, p = 0.001. In LBBB ECGs, intraclass correlation coefficients (ICCs) were comparable for inter-manufacturer and inter-observer agreement (ICC 0.830 versus 0.837). When assessing paced QRSD, manual measurements showed higher ICC compared to inter-manufacturer agreement (ICC 0.902 versus 0.776). Using guideline cutoffs of 130 ms, up to 15% of the LBBB ECGs would be misclassified as < 130 ms or = 130 ms by at least one method. Using a cutoff of 150 ms, this number increases to 33% of ECGs being misclassified. However, by combining LBBB-morphology and QRSD, the number of misclassified ECGs can be decreased by half. Conclusion: Inter-manufacturer differences in computer-calculated QRSD assessments are significant and may compromise adequate selection of individual CRT candidates when using QRSD as sole parameter. Paced QRSD should preferentially be assessed by manual QRSD measurements. (C) 2017 Elsevier B.V. All rights reserved.